Experimental study: brachial motion artifact reduction in the ECG

This study focuses on a dual-input adaptive noise reduction technique by investigation of brachial motion artifact in the ECG under a special experimental protocol. The ECG and motion artifact signals are acquired from a three-electrode system. Primary input is obtained from the standard EGG lead II. Because limbs function like fixed resistors in ECG measurement, we obtain EGG-free brachial motion noise between two electrodes located on the arm, near the right biceps muscle. The separation distance of the electrodes is 5 mm to acquire the motion noise signal, and this signal is the auxiliary input for adaptive filtering. The results show that the LMS algorithm has a very slow rate of convergence. Comparatively, an RLS algorithm converges almost immediately once motion artifact appears and performs satisfactorily in reducing even rapidly varying brachial artifact. It also significantly improves the low-frequency baseline drift. Although the RLS algorithm imposes a large computational burden, a 33-MHz PC486 can execute the algorithm, written in C language, in real time. To prevent the ill-conditioning matrix in the RLS algorithm when the noise is very small, we add white noise to the auxiliary input The experiment shows that this approach can significantly improve the condition of the matrix.